Skip to content
GPU Finder

AWS vs Azure GPU Pricing

Side-by-side GPU cloud pricing comparison with live data and real-time availability.

AWS:20 GPUs
Azure:9 GPUs
Shared GPUs:5
Updated:

AWS is cheaper on 3 of 5 shared GPU models

No data
$6.88/GPU/hrH100 on-demand
on AWS
Visit AWS
Azure H100: $6.98/GPU/hr · AWS and Azure are the two broadest enterprise GPU clouds. AWS has the deeper GPU instance-family list (P and G families from T4 and A10G up to H100/H200/B200 plus Trainium/Inferentia); Azure's N-series VMs cover the same NVIDIA generations for organisations standardised on Microsoft tooling. Neither publishes a public stock API, and both charge for egress.

GPU pricing comparison

Egress Cost Comparison

AWS

$0.09/GB

Azure

$0.087/GB

Exclusive GPUs

Only on AWS (14)

  • A10G
  • B200
  • B300
  • H200
  • Inferentia
  • Inferentia2
  • L4
  • L40S
  • Radeon Pro V520
  • RTX PRO 4500
  • RTX PRO 6000
  • T4g
  • Trainium
  • Virtex UltraScale+ (VU47P)

Only on Azure (4)

  • A10
  • K80
  • M60
  • Radeon MI25

Which Should You Choose?

Choose AWS if...

  • You want the widest GPU instance choice, including AWS's own Trainium and Inferentia accelerators
  • Your data and orchestration already live in S3, EKS, SageMaker or Bedrock
  • You can use Spot or Savings Plans to bring list prices down

Choose Azure if...

  • Your organisation is standardised on Microsoft — Entra ID, Azure ML, AKS, Microsoft 365 contracts
  • You want enterprise agreements that bundle GPU capacity with the rest of your Azure spend
  • You need specific Azure regions or sovereignty options for your workload

Frequently Asked Questions

Is Azure or AWS cheaper for GPUs?

List prices for the same GPU are usually within a similar band and differ by model and region; the live table compares each model both providers price per GPU. Both also offer spot/low-priority capacity and committed discounts that are not shown in list prices, so your contract matters more than the headline.

Which has more GPU instance types, AWS or Azure?

AWS lists more GPU instance families and generations, and is the only one of the two with its own training and inference accelerators (Trainium, Inferentia). Azure's N-series covers the mainstream NVIDIA line-up.

Do AWS or Azure publish live GPU availability?

No. Neither hyperscaler exposes a public stock API, so GPU Finder shows price-only rows for both; capacity is confirmed in each console at launch time.

Does referral status affect this comparison?

No. Affiliate or referral status does not affect rankings, prices, availability, or editorial verdicts on GPU Finder.